Description: (Applicant's abstract) A major goal of the Protein Structure Initiative (Structural Genomics) is to increase the rate of new biological knowledge obtained from protein crystallography, by developing new methods to enhance the efficiency and throughput of structure determination. This proposal addresses several factors limiting the current efficiency. (1) The amount of biological information increases with increasing resolution of the X-ray structure, but full model building and refinement of the highest resolution structures is disproportionately time-consuming. (2) The amount of information from lower resolution structures is limited partly because simplified models are refined to describe these structures. (3) The utility of the database of known structures is limited partly by the reliability of those structures, and the degree to which their interpretation can be automated. This project will develop and distribute computational tools to speed the refinement, improve the quality, and validate the result of protein crystal determinations, with special emphasis on obtaining more information from the increasing number of high resolution structures. Eventual widespread application of these tools will increase the accuracy and information content of a large fraction of all future protein structural models, including those resulting from the Protein Structure Initiative. The key to this work is the use of expanded structural models that include explicit descriptions of atomic anisotropy and modes of molecular vibration. Analysis of near-atomic resolution structures refined with anisotropic displacement parameters will be used to build and maintain a database of statistical properties seen for these parameters in well-refined structures. These properties will be developed as quality control criteria for validating new protein structures. They will also be implemented as restraint targets for structure refinement at lower resolution. This approach should permit incorporation of explicit models for anisotropy into structural models refined at typical resolutions for protein crystallography (about 2A). This, in turn, will make it easier to identify key biological features such as the presence and conformation of bound ligands, and the nature of hinge and inter-domain motions.